Sage – An LLM council that reviews your coding agent's every move

Summary: Sage monitors your coding agent in real time, automatically reviewing its responses, reasoning, and plans using multiple language models to catch issues before implementation. It operates separately from the main agent, maintaining persistent context and enabling multi-turn discussions for thorough critique.

What it does

Sage runs in a separate terminal, cross-checking every coding agent response with other models while reading the codebase, making tool calls, searching the web, and retaining conversation history. It integrates reviews into a continuous thread, preserving context over time.

Who it's for

Developers using coding agents who want automated, persistent, and contextual feedback on their agent’s outputs to reduce bugs and improve plans.

Why it matters

Sage addresses limitations of existing review tools by providing automatic, context-aware, multi-model critique without blocking the main agent or losing conversation history.